From: Peter Zijlstra <peterz@infradead.org>
To: "Li, Aubrey" <aubrey.li@linux.intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>,
Christoph Lameter <cl@linux.com>, Andi Kleen <ak@linux.intel.com>,
Aubrey Li <aubrey.li@intel.com>,
tglx@linutronix.de, len.brown@intel.com, rjw@rjwysocki.net,
tim.c.chen@linux.intel.com, arjan@linux.intel.com,
paulmck@linux.vnet.ibm.com, yang.zhang.wz@gmail.com,
x86@kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH v1 00/11] Create fast idle path for short idle periods
Date: Wed, 12 Jul 2017 10:34:10 +0200 [thread overview]
Message-ID: <20170712083410.ualmvnvzoohyami5@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <496d4921-5768-cd1e-654b-38630b7d2e13@linux.intel.com>
On Wed, Jul 12, 2017 at 12:15:08PM +0800, Li, Aubrey wrote:
> Okay, the difference is that Mike's patch uses a very simple algorithm to make the decision.
No, the difference is that we don't end up with duplication of a metric
ton of code.
It uses the normal idle path, it just makes the NOHZ enter fail.
The condition Mike uses is why that patch never really went anywhere and
needs work.
For the condition I tend to prefer something auto-adjusting vs a tunable
threshold that everybody + dog needs to manually adjust.
So add something that measures the cost of tick_nohz_idle_{enter,exit}()
and base the threshold off of that. Then of course, there's the question
which of the idle estimates to use.
The cpuidle idle estimate includes IRQs, which is important for actual
idle states, but not all interrupts re-enable the tick.
The scheduler idle estimate only considers task activity, which tends to
re-enable the tick.
So the cpuidle estimate is pessimistic in that it'll vastly under
estimate the actual nohz period, while the scheduler estimate will over
estimate. I suspsect the scheduler one is closer to the actual nohz
duration, but this is something we'll have to play with.
next prev parent reply other threads:[~2017-07-12 8:34 UTC|newest]
Thread overview: 114+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-07-10 1:38 Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 01/11] sched/idle: create a fast " Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 02/11] cpuidle: attach cpuidle governor statistics to the per-CPU device Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 03/11] cpuidle: introduce cpuidle governor for idle prediction Aubrey Li
2017-07-12 12:16 ` Peter Zijlstra
2017-07-10 1:38 ` [RFC PATCH v1 04/11] sched/idle: make the fast idle path for short idle periods Aubrey Li
2017-07-11 12:58 ` Paul E. McKenney
2017-07-11 16:33 ` Frederic Weisbecker
2017-07-11 18:11 ` Paul E. McKenney
2017-07-12 3:19 ` Li, Aubrey
2017-07-12 5:03 ` Paul E. McKenney
2017-07-12 5:22 ` Li, Aubrey
2017-07-12 12:19 ` Peter Zijlstra
2017-07-10 1:38 ` [RFC PATCH v1 05/11] cpuidle: update idle statistics before cpuidle governor Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 06/11] timers: keep sleep length updated as needed Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 07/11] cpuidle: make idle residency update more generic Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 08/11] cpuidle: menu: remove reduplicative implementation Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 09/11] cpuidle: menu: feed cpuidle prediction to menu governor Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 10/11] cpuidle: update cpuidle governor when needed Aubrey Li
2017-07-10 1:38 ` [RFC PATCH v1 11/11] sched/idle: Add a tuning knob to allow changing fast idle threshold Aubrey Li
2017-07-10 8:46 ` [RFC PATCH v1 00/11] Create fast idle path for short idle periods Peter Zijlstra
2017-07-10 9:29 ` Wanpeng Li
2017-07-10 13:59 ` Li, Aubrey
2017-07-10 14:46 ` Andi Kleen
2017-07-10 16:42 ` Peter Zijlstra
2017-07-10 17:27 ` Andi Kleen
2017-07-11 4:40 ` Li, Aubrey
2017-07-11 9:41 ` Peter Zijlstra
2017-07-11 16:09 ` Frederic Weisbecker
2017-07-11 16:34 ` Peter Zijlstra
2017-07-11 18:09 ` Paul E. McKenney
2017-07-12 11:54 ` Peter Zijlstra
2017-07-12 15:56 ` Paul E. McKenney
2017-07-12 17:46 ` Peter Zijlstra
2017-07-12 18:53 ` Paul E. McKenney
2017-07-12 19:00 ` Paul E. McKenney
2017-07-19 13:43 ` Frederic Weisbecker
2017-07-19 14:51 ` Paul E. McKenney
2017-07-12 12:22 ` Peter Zijlstra
2017-07-12 15:54 ` Paul E. McKenney
2017-07-12 17:17 ` Peter Zijlstra
2017-07-12 17:57 ` Peter Zijlstra
2017-07-12 18:50 ` Paul E. McKenney
2017-07-12 18:46 ` Paul E. McKenney
2017-07-13 8:34 ` Peter Zijlstra
2017-07-12 4:15 ` Li, Aubrey
2017-07-12 8:34 ` Peter Zijlstra [this message]
2017-07-12 21:32 ` Andi Kleen
2017-07-13 8:36 ` Peter Zijlstra
2017-07-13 14:48 ` Li, Aubrey
2017-07-13 14:53 ` Peter Zijlstra
2017-07-13 15:13 ` Li, Aubrey
2017-07-13 18:28 ` Peter Zijlstra
2017-07-14 3:56 ` Li, Aubrey
2017-07-14 15:38 ` Peter Zijlstra
2017-07-14 15:52 ` Arjan van de Ven
2017-07-14 15:58 ` Peter Zijlstra
2017-07-14 16:03 ` Andi Kleen
2017-07-17 9:21 ` Peter Zijlstra
2017-07-17 13:41 ` Li, Aubrey
2017-07-14 15:53 ` Andi Kleen
2017-07-14 16:06 ` Peter Zijlstra
2017-07-14 16:26 ` Andi Kleen
2017-07-17 19:23 ` Peter Zijlstra
2017-07-17 19:27 ` Arjan van de Ven
2017-07-17 19:46 ` Thomas Gleixner
2017-07-17 19:51 ` Arjan van de Ven
2017-07-17 19:59 ` Thomas Gleixner
2017-07-17 19:48 ` Arjan van de Ven
2017-07-17 19:53 ` Thomas Gleixner
2017-07-17 19:55 ` Arjan van de Ven
2017-07-18 3:23 ` Li, Aubrey
2017-07-18 18:55 ` Peter Zijlstra
2017-07-18 3:14 ` Li, Aubrey
2017-07-18 4:45 ` Andi Kleen
2017-07-18 6:43 ` Thomas Gleixner
2017-07-18 6:56 ` Li, Aubrey
2017-07-18 15:20 ` Paul E. McKenney
2017-07-18 15:29 ` Arjan van de Ven
2017-07-18 16:36 ` Peter Zijlstra
2017-07-18 16:37 ` Arjan van de Ven
2017-07-18 17:05 ` Peter Zijlstra
2017-07-19 5:44 ` Li, Aubrey
2017-07-19 14:48 ` Paul E. McKenney
2017-07-19 15:03 ` Christopher Lameter
2017-07-19 16:54 ` Paul E. McKenney
2017-07-20 1:40 ` Li, Aubrey
2017-07-20 12:50 ` Paul E. McKenney
2017-07-20 13:45 ` Arjan van de Ven
2017-07-20 14:19 ` Peter Zijlstra
2017-07-20 16:02 ` Paul E. McKenney
2017-07-18 16:45 ` Peter Zijlstra
2017-07-18 6:59 ` Andi Kleen
2017-07-18 7:19 ` Thomas Gleixner
2017-07-19 6:12 ` Li, Aubrey
2017-07-19 7:55 ` Thomas Gleixner
2017-07-20 1:56 ` Li, Aubrey
2017-07-20 8:11 ` Thomas Gleixner
2017-07-20 13:48 ` Arjan van de Ven
2017-07-18 7:24 ` Thomas Gleixner
2017-07-18 13:23 ` Peter Zijlstra
2017-07-19 13:43 ` Peter Zijlstra
2017-07-13 15:20 ` Paul E. McKenney
2017-07-14 3:47 ` Li, Aubrey
2017-07-14 4:05 ` Paul E. McKenney
2017-07-17 13:24 ` Li, Aubrey
2017-07-17 13:58 ` Peter Zijlstra
2017-07-17 14:02 ` Li, Aubrey
2017-07-12 12:14 ` Peter Zijlstra
2017-07-11 17:58 ` Christoph Lameter
2017-07-12 2:07 ` Li, Aubrey
2017-07-12 2:35 ` Li, Aubrey
2017-07-12 18:10 ` Peter Zijlstra
2017-07-11 9:02 ` Peter Zijlstra
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170712083410.ualmvnvzoohyami5@hirez.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=ak@linux.intel.com \
--cc=arjan@linux.intel.com \
--cc=aubrey.li@intel.com \
--cc=aubrey.li@linux.intel.com \
--cc=cl@linux.com \
--cc=fweisbec@gmail.com \
--cc=len.brown@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=paulmck@linux.vnet.ibm.com \
--cc=rjw@rjwysocki.net \
--cc=tglx@linutronix.de \
--cc=tim.c.chen@linux.intel.com \
--cc=x86@kernel.org \
--cc=yang.zhang.wz@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®